基于大数据与计算智能的城市雨洪灾害风险评估及响应机制研究

With climate change, the rapid progress of urbanization and information technology, disaster management of urban stormwater is facing more uncertainty and big data's challenge in China. How to effectively express the multi-source and heterogeneous big data of stormwater and scientifically construct the standardized big data resource pool? How to scientifically construct risk assessment index system and model of urban stormwater under big data environment? How to establish the response mechanism of urban stormwater? And so on, these are the key issues that urgently need to be discussed in urban stormwater management. This project is based on the theory and method of big data and computational intelligence to study the risk assessment and response mechanism of urban stormwater. The main contents are as follows: Firstly, the characteristics of urban stormwater big data are analyzed, the stormwater integrated markup language is established, and a virtualized big data resource pool of urban stormwater is constructed. Secondly, the evaluation index system of urban stormwater risk is constructed based on the coupling of computational intelligence and pressure-state-effect-response(PSIR)model, and the deep learning model is improved to establish the stormwater risk assessment model. Then, the response mechanism of urban stormwater including early-warning, decision making, emergency rescue, restoration and reconstruction is discussed. Finally, some countermeasures and suggestions for urban stormwater management are put forward, which are based on the empirical study of typical cities.

随着气候变化、城市化和信息技术的快速发展，我国城市雨洪灾害管理面临着更多不确定性以及大数据的挑战。如何有效地对雨洪灾害多源异构大数据进行科学表达并构建标准化的大数据资源池？如何科学构建大数据环境下城市雨洪灾害风险评估指标体系及模型？如何建立城市雨洪灾害响应机制等，是目前城市雨洪灾害管理迫切需要探讨的关键问题。本项目基于大数据与计算智能的理论和方法开展城市雨洪灾害风险评估及响应机制研究。主要内容包括：分析城市雨洪灾害大数据的特点，建立雨洪灾害集成标记语言，构建虚拟化的城市雨洪灾害大数据资源池；研究基于计算智能与压力-状态-影响-响应(PSIR)耦合的模型构建城市雨洪灾害风险评估指标体系，并对深度学习模型(DLM)进行改进，建立雨洪灾害风险评估模型；探讨包括预警、决策、应急救援、恢复重建等方面的城市雨洪灾害响应机制。最后，结合雨洪典型城市开展实证研究，并提出城市雨洪灾害管理的对策建议。

随着全球气候变化以及我国城市化进程的快速推进，我国城市雨洪灾害频发，其管理面临着更多不确定性以及复杂大数据决策的挑战。项目结合城市雨洪灾害风险管理的特点，建立了基于风险识别、风险评估及风险调控的城市雨洪灾害风险管理框架；基于城市雨洪灾害大数据特征，构建了雨洪灾害集成标记语言（Stormwater Integration Markup Language，SIML）及大数据资源池，实现了洪涝灾害数据的标准化传输和使用。基于灾害系统理论及压力-状态-影响-响应（PSIR）模型，从致灾因子、孕灾环境、承灾体、防灾减灾能力四个方面分析城市雨洪灾害风险的影响因素，构建城市雨洪灾害风险评估指标体系。建立了基于随机森林（RF）确定指标重要性的确权方法，实现对雨洪灾害大数据进行特征提取和特征降维。构建了随机森林与深度信念网络耦合（RF-DBN）的城市雨洪灾害风险评估模型，该模型能实现指标重要性评估及根据提取的重要指标作为深度信念网络（DBN）模型的输入，降低了模型的复杂性并能保证模型精确性。结合南京市开展实证及对策研究，得到了各个区的雨洪风险等级及风险图。并进一步拓展到对江苏省13个地级市、长江三角洲城市群26个城市的雨洪灾害评估。本项研究具有重要的推广和应用价值，理论上丰富和发展了城市雨洪灾害风险管理的理论与方法，应用上为政府部门制定决策，提升城市雨洪灾害的管理水平，降低雨洪灾害风险，具有重要的科学意义和应用价值。